[flang-commits] [flang] [Flang][RISCV] Set vscale_range based off zvl*b (PR #77277)

[Philip Reames via flang-commits]

================
@@ -712,13 +714,55 @@ void CodeGenAction::lowerHLFIRToFIR() {
 static std::optional<std::pair<unsigned, unsigned>>
 getVScaleRange(CompilerInstance &ci) {
   const auto &langOpts = ci.getInvocation().getLangOpts();
-  if (langOpts.VScaleMin || langOpts.VScaleMax)
-    return std::pair<unsigned, unsigned>(
-        langOpts.VScaleMin ? langOpts.VScaleMin : 1, langOpts.VScaleMax);
+  const auto targetOpts = ci.getInvocation().getTargetOpts();
+  const llvm::Triple triple(targetOpts.triple);
+
+  if (triple.isAArch64()) {
+    if (langOpts.VScaleMin || langOpts.VScaleMax)
+      return std::pair<unsigned, unsigned>(
+          langOpts.VScaleMin ? langOpts.VScaleMin : 1, langOpts.VScaleMax);
+
+    std::string featuresStr = ci.getTargetFeatures();
+    if (featuresStr.find("+sve") != std::string::npos)
+      return std::pair<unsigned, unsigned>(1, 16);
+  } else if (triple.isRISCV()) {
----------------
preames wrote:

Minor, but if we're going to making this fully conditional on target, go ahead and duplicate the 
  return std::nullopt; and use early return pattern here instead of the else-if chain.  

https://github.com/llvm/llvm-project/pull/77277